n8n-ai-workflows

n8n-ai-workflows is a repository of ready-to-import n8n workflows for AI automation. It includes a Telegram support bot that runs without public HTTPS via a long-polling bridge, a PDF-to-pgvector ingestion pipeline, a Telegram RAG query workflow, a universal site chatbot, and a single-node Document Ingest pipeline using the custom community package n8n-nodes-doc-ingest. The project ships with a Docker stack (n8n, postgres/pgvector, qdrant) and a production-ready Telegram poller.

The repository addresses five common problems: running a Telegram bot without public HTTPS, reliable delivery without duplicates, private-document RAG, universal site chat without per-site forks, and single-node ingestion with a custom package. For example, it uses poller.py to long-poll getUpdates and forward updates to a local Webhook node, allowing operation on localhost, behind CGNAT, or on IPv6-only hosts without ngrok or a domain. It also provides a one-shot ingestion pipeline that reads local PDFs, extracts text, chunks it, embeds with text-embedding-3-small, and inserts into the embeddings table.

The workflows include customer-support-agent (Telegram AI Support Bot), rag-doc-search (RAG over PGVector with ingest and query), lead-scoring-crm, universal-site-agent, and doc-ingest-custom. Each workflow is validated with python3 -m json.tool and uses @n8n/n8n-nodes-langchain.* types where required.